Global Water withdrawal Analysis¶
This research analyses global water usage patterns across different countries and time.
The water withdrawal is divided into three main uses agricultural, domestic, and industrial and is going to be compared with GDP and population.
The idea came from the Global Water Distribution Infographic created by Chesca Kirkland. The topic warrants comprehensive analysis especially since Financial analysts like Michael Burry thinks that water scarcity is one of the topics that will drive future economies (Killik&Co (ed), 2017).
Due to this relevance, many visualisation projects have been published, Our World in Data and Worldometer dataset are two of the main websites that provide open source datasets that have also been used for this research (Ritchie & Roser, 2018).
The dataset used for this research comes from the interpolation of many different open datasets from World Bank Open Data, Worldometer dataset, Our World in Data, and from studies from World Water Organisation to obtain geospatial and temporal analysis.
Water Usage Distribution by Sector¶
In order to observe geographical trends of water usage changes within each country, the following analysis will show the distribution across agricultural, domestic, and industrial sectors (green, red and blue respectively; this color coding will be consistent across the whole research). The map uses RGB colour mixing to represent the proportion of each type of usage.
For this plot, the data was gathered from the article Freshwater Withdrawal by Country and Sector (2013 Update) and was then converted into a csv with Zamzar. This type of mapping can visually show geographical trends on water usage proportions. To do that, each country’s water usage distribution is represented as an RGB tuple, if a country uses mostly agricultural water, its colour will be reddish, and the combinations of different values will result in mixed colours.
As the plot shows, the north of the world have higher industrial water usage in proportion and low agriculture water use while warmer countries have a proportional higher agricultural withdrawal.
Thanks to this visual representation is possible to observe how geographical location may influence water usage proportions.
Agricultural vs Industrial Water Withdrawal with GDP Comparison¶
Inspired by the data visualisation work by Hans Rosling, with the following plot is possible to appreciate the relationship between agricultural and industrial water withdrawal across different countries, with GDP as a sizing factor. This animated visualization shows how these relationships have evolved over time from 1965 to 2020. China, India and USA have been removed from this plot since they skewed the result due to dimension of their economies. Missing data have been linearly interpolated and since original records had a 5 years interval, the slider has a 5 year interval too.
Is possible to inspect each continent by double-clicking it in the legend, to evaluate specific grouped behaviours.
One insight to appreciate is that with the increase of GDP, countries stabilise in the proportion of water use.
Total Water Withdrawal per Capita Over Time¶
The following plot will help to understand how water consumption patterns have changed globally. Missing values have been interpolated.
After having adapted the data to the map, an animated choropleth map has been used to visualize how water withdrawal per capita has changed over time.
Dividing total water withdrawal by the population is useful to normalise the data but states like Turkmenistan that have a low population and high water consumption resulted as an outlier with 5.6 thousand m^3 of water used per capita. The colour grading of this plot stops at 2500 m^3 so that the rest of the countries have a more visible outcome.
The trend of water consumption for several countries has followed a distinctive trajectory with an initial increase (probably followed by an increase of economy and living standard improvements), followed by a decline in recent years, likely driven by water efficiency advancements or polices. Other countries may not have yet reached their peak on water consumption but is possible to hope that more efficient water management strategies will be implemented.
Water Consumption and GDP by Continent¶
The following boxplot explore the variation and distribution per capita of water consumption and GDP across all continents. The water consumption has been multiplied by 0.05 so that the variables had same mean to make the boxplots comparable.
From this plot is clear that there is a general trend, as continents like Africa have both low per capita water use and GDP. While continents with higher GDP (i.e. Asia) have an overall higher consumption of water. Europe in this case is a positve example of decoupling between water use and GDP, while continents like South America present an high per capita water consumption while having a low GDP.
Overall GDP vs water consumption¶
The following scatterplot is used to explore the above mentioned possible correlation between GDP and water consumption for all the countries without a continent distinction.
The scatterplot present a local regression trendline (locally weighted scatterplot smoothing), that appeared in statistical modelling in 1979, and is used to show the correlation between these two features. The LOWESS regression has been chosen to capture non-linear trends and highlight local patterns. Indeed low GDP countries behave differently from high GDP countries for water use fluctuations; and thanks to this regression is possible to compare trends across countries effectively. The LOWESS regression also helps to avoid overfitting by capturing patterns and by smoothing out noise or outliers (CLEVELAN, 1979).
In this plot is possible to see the point of inflection of the curve that is around 4000 of GDP, from that point the inclination decreases, so the variation of GDP has less influence on the variation of per-capita water use.
The plot can be inspected by excluding each continent points by clicking them on the legend. To explore more in detail countries with lower GDP per capita is possible to select an area to zoom it. This output seem to be connected to the studies of the economist Kuznets who, studying the homonymous u-shaped curve, shows the relation between economic growth and environmental quality. According to Kuznets, economic growth brings to lower environmental quality, income levels and higher deforestation only up to a certain point, then the technological progresses bring to a decrease and this might be an explanation to the inflection of the regression line (Shahbaz et al., 2023).
Correlation values between GDP and specific water uses¶
The correlation plot can explain the magnitude of the correlation between types of water (Domestic, Industrial, Agricultural) use and GDP.
For this case the variables have all been logarithmically transformed since the data is heavily skewed.
Thanks to this plot is possible to give a value to these correlations and is clear that industrial and domestic water use have a positive correlation with GDP while the agricultural doesn’t.
Is also clear a correlation between industrial and a Agricoltural use.
To explore how these variables behave a line-graph is shown below.
The variables are log-transformed to facilitate the visualisation of the relationships across the scales, since GDP and water usage have different magnitudes, then the countries were ordered in ascending GDP values.
By clicking the “Agricultural Consumption” label is possible to see that even if the data varies, there is an actual increase between per capita GDP and per capita domestic and industrial use.
Since this data is log-scaled is not possible to appreciate the proportion on the water withdrawal, To do that the first choropleth chart can be reviewed, and to appreciate the proportions across the continents and over time, the following graphs will help.
Water Usage, GDP, and Total Water Withdrawal by Continent¶
Through the use of a stacked bar plots it is possible to compare the usage proportion between each continent. Since the height of each bar is constant for comparative reasons, a line graph is added to show the total water withdrawal of each continent.
A second trend line is used to appreciate the cumulative GDP level of each continent.
Excludiing Europe and ASIA, is seems that Total water withdrawal and total GDP are correlated and that continets like North/Central America and Europe have a higher industrial consumption while larger territories like Asia and South America have higher proportional water consumption.
While Domestic use seem to change less across continents.
Proportions of Water Usage Over Time¶
A stacked area-chart can show how proportion of water usage changes over time, and with the use of Plotly, is also possible to see individual trends of domestic, industrial and agricultural withdrawals.
For this graph 2020 data has been omitted due to missing values.
By watching the graph is clear that while total water usage over time is increasing, after 2010 there was a decrease, that can be driven by the decrease of total water use of biggest countries which corroborates the results of the second choropleth chart (Water withdrawal per capita over time).
This graph helps analyze data up to 2015 but to predict future trends, monthly data and more variables would be needed, because innovation is yet bringing to a rapid decrease of water consumption. For exmple agriculture is already adopting water-efficient methods like drip irrigation and rotational grazing (DGB Group, 2024).
Given these inputs is fair to guess why total water use has decreased and is possible to assume that water waste might decrease if institutions will invest in minimising water withdrawal.
Water Consumption Flow Analysis (Sankey Diagram)¶
Finally, a Sankey diagram visualises the flow of water consumption from countries to continents to usage types (agricultural, industrial, and domestic). All countries have been kept to show both individual use and continental distribution of water consumption. Aggregating countries in macro-regions might have brought to less information gained from the plot and Plotly’s interactive features allows zooming and selection.
The nodes sizes shows the consumption of each country, continent and process, and proportional relationships between countries and water use.
Agriculture dominates global water use, shown by the thick green flow ribbons connecting it, especially in Asia. Industrial use is the second largest use with Europe significantly contributing. Asia stands out as the region with the highest overall water consumption due to its many countries and by countries like Turkmenistan with the highest overall per capita water use.
Like in the above boxplot chart (Water Consumption and GDP by Continent) is visible the country variation in water use within the continents, reflecting possible differences in economic developement, climate and economy structure.
Such a graph may be used by policymakers to identify where conservation efforts in water usage might have the greatest impact.
Data Analysis Aspects investigated¶
In this research different data analysis techniques were used.
Correlation analysis, done with the corr-plot, the linegraph and the stacked barplot with trend-lines suggest a corelation between GDP and total water use with a higher correlation between industrial water use and GDP.
Temporal analysis has been shown with the stacked area-chart and the two animated plots that showed an initial increase and a following decrease of water use over time.
Thanks to the boxplots, the Sankey diagram and the stacked bar chart it was possible to visualise a comparative distribution analysis that showed how water use and GDP differs across countries and continents with different variances and amounts of water withdrawal.
The choropleth graphs were useful for a spatial pattern recognition analysis, with the first one that gave insights on geographical patterns on water use and the second one on the evolution on per capita water usage over time.
Future Work¶
This reseach is a significant starting point for possible future analysis which could include other variables like cost of water across different regions and in different time periods. Another correlation to be explored might be the effect of different diets across the world on water usage.
Exploring these trends might be useful to create a model to estimate future water consumption that can be compared with total water resources available, such a study may be interesting to evaluate future fresh water prices and when it will be economically convenient to start using plants for desalination of seawater for agriculture and industry uses.
At the moment there are many organisations that are exploring trends and possible solutions to water waste like the World Water Council or The World’s Water by the Pacific Institute, which gather data and try to inform policy makers and public.
A more extensive research could be an important piece in the mosaic of information on water consumption.
Bibliography¶
Data sources¶
- Worldometers - Water
- Our World in Data - Water Use Stress
- Our World in Data - Water Data
- World Bank - Population
- World Bank - GDP
- World Water Report (PDF)
- World data over time
- Mapping data
Code Sources¶
- seaborn choropleth chart
- Plotly choropleth map
- stacked barplot with lines
- Trend lines
- Sankey diagram, Sankey diagram-theory
- Corrplot
- Animated plots
- boxplots
- Area chart
- Hover Text
Bibliography¶
Bhandawatchetanya (2023) Virtual water infographic - virtualisation analysis, Medium. Available at: https://medium.com/@bhandawatchetanya/virtual-water-infographic-virtualisation-analysis-eb3734c25e26 (Accessed: 07 March 2025).
CLEVELAN, W.S. (1979) Robust Locally Weighted Regression and Smoothing Scatterplo, Statistical Association. Available at: https://home.engineering.iastate.edu/~shermanp/STAT447/Lectures/Cleveland%20paper.pdf (Accessed: 08 March 2025).
Editor (2024) 10 agricultural techniques for water conservation, DGB Group. Available at: https://www.green.earth/blog/10-agricultural-techniques-for-water-conservation (Accessed: 08 March 2025).
Killik&Co (ed.) (2017) Why Michael Burry of 'The big short' is investing in water, Why Michael Burry of 'The Big Short' is Investing in Water Killik & Co. Available at: https://killik.com/articles/why-michael-burry-of-the-big-short-is-investing-in-water/ (Accessed: 07 March 2025).
Ritchie, H. and Roser, M. (2018) Water use and stress, Our World in Data. Available at: https://ourworldindata.org/water-use-stress (Accessed: 07 March 2025).
Shahbaz, M. et al. (2023) Environmental kuznets curve, Environmental Kuznets Curve - an overview | ScienceDirect Topics. Available at: https://www.sciencedirect.com/topics/earth-and-planetary-sciences/environmental-kuznets-curve (Accessed: 08 March 2025).